Swift: avoid module scan for lazy body emission

[arun279]

Summary

Avoid enumerating every top-level declaration in a Swift module when the extractor is processing one lazy declaration. extractLazy passes the pending declaration's own module, and shouldEmitDeclBody separately rejects every declaration except that exact currentLazyDeclaration, so populating currentTopLevelDecls cannot affect the result.

Performance

I profiled a reproducible CodeQL scan of a small Xcode project at 10 ms intervals. Of 33,602 samples in lazy extraction, 28,176 were in ModuleDecl::getTopLevelDecls while constructing SwiftBodyEmissionStrategy.

On identical source, runner, Xcode, CodeQL 2.26.3, architecture, and build flags:

Measurement	Unpatched	Patched
Linked workflow wall time shown by GitHub	27m37s	28m55s
Benchmark-only extractor source build	skipped	9m33s
Build analyzed application step	22m03s	11m53s
Resolve Package Graph	8m46s	4m42s
Helper-to-app compile interval	10m33s	4m02s

The patched workflow is 1m18s longer overall because this diagnostic cell first builds the modified extractor from source. That setup is absent from the unpatched cell and would not exist when the extractor ships prebuilt with CodeQL. The measured claim here is a 46.1% reduction in the analyzed-application build step, not an end-to-end reduction in these two diagnostic workflows.

• Unpatched run
• Patched run
• Related package-resolution report: Swift build-mode: manual — xcodebuild "Resolve Package Graph" hangs ~12 min under the build tracer (local path packages, nothing to fetch) #22121

Correctness checks

Both runs completed analysis with:

• the same 22 extracted project source files;
• the same 37 sorted SARIF artifact URIs;
• 1,892 successfully extracted project lines;
• zero unresolved AST nodes;
• zero findings.

The patched database was not smaller: its Swift AST-node metric was 699,967 versus 696,067 and its TRAP data was 585.74 MiB versus 583.35 MiB. Both source archives contained the same 22 files with byte-identical contents. No extraction-quality regression was observed.

The current CodeQL branch passes git diff --check and clang-format --dry-run --Werror. The patched extractor was also compiled from source and exercised by the patched benchmark run. At this PR's base commit, SwiftExtractor.cpp and SwiftBodyEmissionStrategy.{h,cpp} are unchanged from CodeQL 2.26.3.

[jketema]

This looks correct. Thanks for your contribution!

[Copilot]

Pull request overview

Optimizes Swift lazy declaration extraction by avoiding an unnecessary module-wide declaration scan.

Changes:

• Skips populating currentTopLevelDecls when extracting one lazy declaration.
• Preserves existing body-emission filtering behavior.

Show a summary per file

File	Description
swift/extractor/infra/SwiftBodyEmissionStrategy.h	Avoids the costly top-level declaration scan during lazy extraction.

Review details

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Balanced